Data processing: speech signal processing – linguistics – language – Speech signal processing – Application
Reexamination Certificate
1998-07-29
2002-02-12
Dorvil, Richemond (Department: 2641)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Application
C704S271000, C704S275000
Reexamination Certificate
active
06347299
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention comprises a human-computer interface for making spoken inputs to an automatic speech recognition system and receiving audio outputs from the computer system in a manner that enables creation, access, and modification of electronic records without visual display of the records and without use of hand-controlled devices, such as keyboard, touch screen, mouse, or other pointing device.
2. Description of the Prior Art
Prior art systems have been developed that allow computers to record and recognize human speech, converting the sound into text or allowing machines to perform different activities in response to different spoken commands. For example, some telephone applications allow users to make selections from auditorially presented lists of choices by saying the number of the choice. However, computers often fail to correctly recognize words and phrases, especially when the allowable vocabulary increases beyond a small set of words (e.g., single digit numbers), when the system is not tailored to the voice and pronunciation of one particular speaker, and when background noise is loud or variable. Also, users themselves sometimes say the wrong thing and wish to correct it, or they may wish to update the contents of a record with a new value. Therefore, for speech recognition to be a practical means of making inputs to electronic records, there must be some means by which the user can obtain feedback about how the computer interpreted the inputs and to correct inputs not properly recognized by the computer, incorrectly input by the user, or in need of updating by the user. The prior art does not provide an efficient mechanism for obtaining feedback about inputs made through speech and for modifying the record without the use of a visual display or manual entries (e.g., made with a keyboard or mouse).
If inputs and queries are to be performed with electronic records without either a display or manual entry mechanism but with the flexibility that accompanies conventional visual-manual entry methods, there must be a way to direct inputs to specific fields in the record, and query the contents of fields. The prior state of the art does not sufficiently address this need. In the prior art, such as interacting with telephone menu systems, fields are typically filled by single word responses made to audio prompts that are provided in a predetermined sequence.
Computer-generated sounds, such as tones, waveforms, and speech have been on the market for years. For example, virtually all personal computers have the ability to emit distinctive tones to signal such things as the successful startup of the operating system, exit from the operating system, and attempts to perform invalid computer operations, among other things. Likewise, there are systems that take typed text as inputs and produce outputs that sound similar to humans speaking the typed inputs. However, these prior systems do not use audio to help a user navigate through an electronic record in combination with a speech recognition system as described in this invention.
SUMMARY OF THE INVENTION
The present invention provides speech and audio user-computer interface mechanisms for accessing and editing information in electronic records. A mechanism is provided by which the user can direct inputs to any of a variety of fields without following a predetermined order of input. This allows the user to be proactive in making entries rather than simply reacting to requirements set by computer-generated prompts.
In the present invention, audio is provided as feedback to the user, not as a fixed path prompt for the user. This feedback can be in the form of non-verbal auditory signals or synthesized speech. The invention uses audio to inform the user of whether or not the system understood the spoken words or phrases as valid inputs to the electronic record, what the system recognized as the input, and to identify the contents of various fields in the electronic record. The precise wording for the speech inputs can be changed from one implementation of the invention to another, depending on what terminology is most meaningful to users, works best with the speech recognition engine being used, etc. Likewise, the audio outputs from the system, both nonverbal sounds and synthesized speech, used in implementing this invention can vary from one application to another. An illustrative set of speech inputs and auditory outputs that could be used to implement this invention is provided.
The present invention is designed to work with electronic records used by interactive computer applications for which fields (and sets of fields, in the case of hierarchically organized records) are defined in advance and names or labels are assigned to those fields (and sets of fields) to make them identifiable through speech. These records are assumed to reside in an electronic storage medium that can be read and accessed by the computer's central processing unit.
Users create or initiate a new electronic record by issuing a spoken command such as “New record.” This ensures that a computer file with the predefined record structure is opened so that the user can begin speaking inputs to its fields. Users can work with multiple electronic records through this technique. To distinguish one record from another, the system can assign unique identifiers (e.g., identification numbers) to the record by default, and the user can speak identifying information for the appropriate field(s) of the record (e.g., input the name of the object for which the record was created) as well.
The user can open any record accessible by the application implementing this invention by saying “Open record” followed by any previously assigned identifier for the record. Likewise, a record can be closed by saying “Close record” followed by any appropriate record identifier.
Since the present invention enables users to work with electronic records without being able to see an identifier for the record, a mechanism is provided for querying the computer by speech for the identifier assigned to the record. The computer responds with synthesized speech output of the record identifier. This helps the user to keep track of which record upon which is being worked, which the user may have forgotten or may not have known because of the manner in which the record was opened.
The present invention provides means for more than one electronic record to be “open” at once. It also provides rules and speech commands for governing which of these records is “active” (ready to receive inputs and provide outputs) at any given time.
Distinctive nonverbal auditory signals are included in the invention design to indicate to the user whether or not the system understood a spoken input to the record to be valid. Inputs signaled as invalid by a brief clicking sound are not committed to the record. Spoken inputs interpreted as valid data entries are signaled as such by a brief chime and are committed to the electronic record.
Users can monitor what the system interpreted as their input and committed to the electronic record by requesting synthesized speech feedback of what the system understood the user's input to be. This feature of the invention can be activated by a speech command such as “Start speech feedback” and deactivated by a speech command such as “End speech feedback.”
Users can request output of the entire contents of the electronic record by issuing the command “Play back log.” The system responds with a synthesized speech readout of the name, or label, and contents of each field. The present invention also enables a user to request the contents of a particular field by saying “Play back” followed by the name of the field. The system then responds with synthesized speech output of the contents of the requested field. The user can also request the contents of a particular predefined subset of fields by saying “Play back” followed by the name of the subset. The system responds by saying the label of each field in the requested
Amin Dick S.
Holzman Thomas G.
Huff Monica M.
Hunter Wesley G.
Jacobson Mark A.
Dorvil Richemond
McFadden Susan
NCR Corporation
Needle & Rosenberg P.C.
LandOfFree
System for navigation and editing of electronic records... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System for navigation and editing of electronic records..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System for navigation and editing of electronic records... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2954950